Pavement Deterioration Modeling and Prediction for Kentucky Interstate and Highways

Pavement management and preservation (PMP) is as important to our nation’s highway infrastructure development as the construction of new infrastructures. On the other hand, in order for transportation authorities to properly allocate resources and prioritize among the PMP projects, a prediction model for pavement deterioration is necessary. This paper studies the pavement deterioration for Kentucky interstate and highways using statistical and data mining methods. Two models of linear regression (LR) and artificial neural networks (ANN) are developed to predict the deterioration of wheel path cracking (WPC) over one year period. Particularly, two indices on WPC, i.e., extent and severity of WPC, are target/output variables, while the two WPC indices of the current year, age and average daily traffic are input variables. Original data includes measurements on 5,146 road segments over 11 years. Efforts on data preprocessing, input analysis, model testing and validation, as well as comparisons of the two models are reported. Results from SAS Enterprise Miner 12.1 suggest that both methods produce comparable and quality prediction, with the average squared errors over three data subsets (training set of size 645, testing set of size 322 and validation set of size 322) around 1.0.